mne.compute_rank does not return information for all valid channel types

[tsbinns]

Description of the problem

mne.compute_rank calls mne.io.pick._picks_by_type to find the channels to compute the rank for.

_picks_by_type only returns channels of types given in the variable _DATA_CH_TYPES_SPLIT which does not include types such as 'misc', 'emg', etc..., meaning information is not returned for all valid channel types. https://github.com/mne-tools/mne-python/blob/3bf04edf744da89b1bb4f84cde5537758ed00865/mne/io/pick.py#L837-L862 https://github.com/mne-tools/mne-python/blob/3bf04edf744da89b1bb4f84cde5537758ed00865/mne/io/pick.py#L939-L943

If this behaviour is intended, perhaps it could at least be noted in the documentation of compute_rank that information will not be returned for certain channel types, or raised as a warning message that channels of valid types not marked as 'bads' were dropped.

Steps to reproduce

import numpy as np
import mne

rand = np.random.RandomState(44)

n_channels = 3
n_times = 50
data = rand.random((n_channels, n_times))

info = mne.create_info(
    ch_names=[str(i) for i in range(n_channels)],
    sfreq=256,
    ch_types=['misc' for _ in range(n_channels)])

raw = mne.io.RawArray(data, info)

rank_dict = mne.compute_rank(raw)
rank_sum = sum(rank_dict.values())

assert np.linalg.matrix_rank(data) == n_channels  # expected rank (3)

Link to data

No response

Expected results

rank_dict == {'misc': 3} rank_sum == 3

Actual results

rank_dict == {} rank_sum = 0

Additional information

Platform: Windows-10-10.0.22621-SP0 Python: 3.11.0 | packaged by conda-forge | (main, Oct 25 2022, 06:12:32) [MSC v.1929 64 bit (AMD64)] Executable: C:\Users\User\anaconda3\envs\mne_ssd\python.exe CPU: Intel64 Family 6 Model 140 Stepping 1, GenuineIntel: 8 cores Memory: 47.7 GB

mne: 1.4.dev0 numpy: 1.24.2 {MKL 2022.1-Product with 4 threads} scipy: 1.10.0 matplotlib: 3.6.3 {backend=TkAgg}

sklearn: 1.2.1 numba: Not found nibabel: Not found nilearn: Not found dipy: Not found openmeeg: Not found cupy: Not found pandas: 1.5.3 pyvista: 0.38.1 {OpenGL 4.5.0 - Build 31.0.101.3889 via Intel(R) Iris(R) Xe Graphics} pyvistaqt: 0.9.1 ipyvtklink: Not found vtk: 9.2.5 qtpy: Not found ipympl: Not found pyqtgraph: Not found pooch: v1.6.0

mne_bids: Not found mne_nirs: Not found mne_features: Not found mne_qt_browser: Not found mne_connectivity: Not found mne_icalabel: Not found

[welcome[bot]]

Hello! 👋 Thanks for opening your first issue here! ❤️ We will try to get back to you soon. 🚴🏽‍♂️

[larsoner]

Yes I think we should document this better, only channels considered data channels should be included by default

[agramfort]

I believe now that EMG is data :)

A

Message ID: @.***>